Slurm script fails when I request mpirun/ntasks

[slabhan1]

I am trying to run a LIS (module openmpi_5.0.0) with the following slurm script on Milan:

#!/bin/bash

#SBATCH -J cold_30node
#SBATCH --constraint=mil 
#SBATCH --ntasks=1

#SBATCH --time=00:00:40
#SBATCH --output=./LDAS_irri/lisjob.out
#SBATCH -D ./wd/
#SBATCH --no-requeue
#SBATCH --mail-type=ALL
#SBATCH --error=./LDAS_irri/error.log
module purge
source $HOME/.profile


module use --append $HOME/privatemodules

module load lisf_7.5_intel_2023.2.1_openmpi_5.0.0

time mpirun -np 1 ./LIS -f ./LDAS_irri/lis_config_2001

exit 0

Which throws the error:

I have tried to run this without "mpirun -np 1" and "ntasks = 1" and it works just fine, so I imagine it's not an issue with the executable/config. I can't figure out why I can't run this with mpirun, please help!
mpirun --version
mpirun (Open MPI) 5.0.0

sinfo --version
slurm 22.05.11

[dmocko]

Hi @slabhan1

Sorry for the delay in responding. If you are only running with 1 task, you can just run with this command:

./LIS -f ./LDAS_irri/lis_config_2001

It sounds like this is already working for you. I think "mpirun" can cause errors when running with just 1 task.

If that's not the issue, are you sure that you have both compiled and submitted the job from a Milan node?

I see you are on Discover. You can point me to your LISF/lis/make/configure.lis (ensure permissions are open) and I'll take a look.

I just posted a separate discussion. Please see if Issue 5 in it might apply to you. #1642

My apologies, I see that we never responded to your question from March: #1503

Would you still like us to respond to that other question?